vanadium-browser-test-web is stuck on Initializing Vanadium

[afandria]

This test fails about 25% of the time on Jenkins. According to the screenshots, we get stuck on initializing Vanadium, so this may be a timing issue. It's possible that my test's timeout is too short.

For now, I have disabled the test. I plan to take a look on Tuesday.

@arup42 @aghassemi @wmleler

[afandria]

Re-enabled with a longer timeout. Has not failed for the past few runs, but will wait for a day to see if the issue has really been fixed.

[arup42]

We have a timeout in this test again. Is this different from the earlier timeout?

org.openqa.selenium.TimeoutException: Timed out after 20 seconds waiting for io.v.webdriver.nsbrowser.MainPage$1@3d894c74

[afandria]

Same problem. The caveats tab is not appearing after the page load. (Note: This also affects the other web tests.)

20s (and even 10s, the previous timeout) are really a long time to wait for the tab to show up, and it doesn't fix the problem. I will have to try something else.

[arup42]

Any ETA on this one? (Can we do something to make the test green in the meantime? The last 6 runs have failed.)

[afandria]

No ETA yet, I still have not identified the problem. Something in the extension is hanging, and while it may be something between wspr and js, it may also be that the identity server isn't responding (and the test timeouts before the identity server gets back to us.)

Even more oddly, the tutorial tests do approximately the same thing, and they have a much lower rate of failure.

For now, I've disabled the test.

[afandria]

The current plan is to see if the identity server is at fault.

• add log lines to prod identityd @asimshankar
• add a console log of the public key in the JS code. (Is it possible to obtain this during vanadium.init?)
• obtain the ChromeDriver console output in the webdriver test and ensure that it is available on the Jenkins console output.

Analysis of the ingoing/outgoing log lines and timestamps may implicate/absolve identityd.

[arup42]

@jakline

Note that Alex's theory is that disabling this test caused the problem to move to the vanadium-www-tutorials-js-web test. There's a separate top-level bug open for that one:

vanadium/issues#520

[afandria]

A more short-term solution (which doesn't actually fix the underlying bug) would be for the test to try again if it encounters this problem. I will likely end up implementing this so that the test can be re-enabled. Since the test would try multiple times, we can still succeed while collecting debug information.

[afandria]

The retry-logic has been added as well as more screenshot and console log grabbing. Once we get another failure, I'll start inspecting logs. (Note: The tests may show up as green because of retry logic. I will have to look at the console output and screenshots to determine if the problem occurred.)

[afandria]

This test is fine now with the retry logic. The theory that this test "warms up" the identity server seems to be correct, though the underlying cause for its sql server's behavior is still unknown. The bug with identityd is being tracked separately.